Summary: Google published a Dialogues on Technology and Society episode on September 14, 2026 with astronaut Christina Koch and research executive James Manyika. The summary says Koch drew on 328 days in space, the first all-female spacewalk, and Artemis II while discussing exploration and the partnership among astronauts, robotics, and AI. It frames exploration as a division of labor between human judgment and automation.
Ask most people where AI belongs in exploration and the answer starts with what the model can do. The Dialogues on Technology and Society episode published September 14, 2026 opens on a more useful question: what does a person bring, and what does a machine bring, in a place neither of them reaches alone.
Who is sitting down
NASA astronaut, engineer, and scientist Christina Koch appears with James Manyika, Google’s Senior Vice President of Research, Labs, Technology & Society. On Koch’s record: 328 days aboard the International Space Station, the first all-female spacewalk, and a trip around the Moon as part of NASA’s Artemis II mission.
Per the supplied RSS summary, the conversation ranges over Earth seen as an electric blue “lifeboat” from 250,000 miles away, the partnership among astronauts, robotics, and AI, the question “Are we alone?”, and her advice to future explorers: do what scares you, and support others.
The partnership framing is the transferable part
Describing astronauts, robotics, and AI as a partnership is not decoration. It is an operating model, and it carries over directly to how teams should reason about agents.
Koch’s work does not fit the pattern of a person supervising a tool that could have handled the job on its own. She is the one exercising judgment where a wrong call costs the most, while automation covers what people are poor at: precision that never drifts, procedure that repeats identically, and monitoring that does not tire.
That is a different brief from “replace the manual step.” It produces two lists: decisions that need a person in the loop because the context is ambiguous, and decisions that go to a machine because the context is monotonous. For anyone building agent workflows, that split is the architecture itself. The same question surfaces in execution models for orchestrators, where who holds state determines how far you can trust the system when something breaks.
“Are we alone?” is a long-horizon search problem
The summary says Koch takes up that question in the episode. It does not say how she or Manyika frame it, so treat the framing as something to watch rather than something to paraphrase.
What the summary does establish is that the question sits alongside the robotics-and-AI partnership discussion. That adjacency is the interesting part for anyone building research or discovery tools: generating a candidate answer is rarely where the difficulty lies. The difficulty is agreeing on what counts as evidence, and staying honest about how little of it you have.
What a builder can act on
Two practical takeaways, both modest.
First, if your product puts a person beside an automated system, write the division down. Which calls require human judgment because the stakes or the ambiguity run high, and which are handed off because consistency matters more than nuance? The episode’s framing treats that division as the design, not as an implementation detail.
Second, watch the episode if you want the primary source. The summary stays thin on specifics — it does not detail the AI systems discussed, the mission timeline, or Manyika’s positions. The supplied RSS summary does not specify those, and that is not evidence they were left out of the conversation.
Koch’s advice as the summary reports it — do what scares you, and support others — is the least technical line in the piece and probably the most reusable. Exploration programs work because people take on work at the edge of their competence with support around them. Teams shipping AI tools are running a smaller version of the same thing.